Abstract: The purpose of image registration is to remove inconsistencies in geometry between the matched registration image and the reference image,including translation,rotation and scaling transform,which is the necessary premise for image contrast,data fusion,transform analysis and target recognition.A new kind of remote sensing image registration,which is called Fourier-Mellin transform method is put forward.Firstly,Fourier transform is performed to get frequency spectrum power for the reference image and the matched image respectively.Next,high-pass filtering is conducted with regard to their energy spectrum.Then,energy after filtering is converted into log-polar coordinate formation.In the case,mutual power spectrum can be computed adopting phase correlation technique to obtain their rotation angle and scale factor correspondingly.Lastly,transform image may be gotten by these parameters.And again,for reference image,high-pass filtering is performed to reduce background noise and frequency aliasing interference in the transformation process.Then,phase correlation calculation may achieve translation value,another transform image can be gotten through translation transform.Finally,registration image may gain by the two images' superposition.The experimental results show that the algorithm is efficiency and reliability.